TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Judgment · AGE 16—18 · POSITIONAL · HELD
The machine did not invent the courtyard
Your region's passive moves come first, and the simulation tests them on this site.
The dilemma
Two schemes for a house in Vijayapura. One came from an image tool briefed "sustainable, eco, net-zero": a sculpted roof, a green wall, solar panels at an angle that looks right. The other is a plain courtyard house with deep verandahs and thick walls — the type the town has built for centuries. The jury is in an hour. The first looks like the future. The second looks like your grandmother's house. Which do you pin up?
The choices
Pin the generated scheme and argue for it from its features. Pin the courtyard and argue from the climate. Or do what a serious office does: take the passive moves of the region first — orientation, shade, mass, cross-ventilation, a court open to the sky — check them against this site with a weather file and a simple simulation, and only then ask whether anything in the generated scheme adds a move the courtyard lacks.
The consequence
Argue from features and you are defending a pattern you cannot explain. When the juror asks what the green wall does in a Vijayapura May, all you can do is describe the picture. Argue from climate and every element has a reason a juror can test, and a number from the simulation to test it with: the court loses its heat to the night sky, the verandah stops the sun before it reaches the wall, the thick wall slows the afternoon heat until the cooler night. Passive first does not forbid new ideas. It asks each new idea to prove it adds something the courtyard does not already do — and it asks the courtyard to prove itself on this plot too.
The case
A third-year student generates twelve "eco houses" for a plot in Belagavi. Eleven have green walls; none has a court. The tutor hands her Correa's writing on open-to-sky space and Baker's drawings of a rat-trap bond wall — bricks laid on edge with a cavity inside, fewer bricks and better insulation than solid brickwork. She rebuilds the scheme around a court and that wall, then runs the image tool again, briefed with those moves and nothing about "eco." The thirteenth image is the first one that would work in Belagavi.
Our position
Passive first. Before any generated scheme, the design answers the climate with the moves its own region has tested for generations — in the Deccan the court, the deep verandah and the thick wall; in Kerala the sloping tiled roof, the nalukettu's inner court and the plan that lets the wind through; in Rajasthan the haveli's jaali and shaded chowk; in Ladakh the thick mud wall and the south-facing glazed room; in Assam the light, stilted house above the flood. There is no one Indian repertoire. There is the one your site belongs to. Then the simulation: precedent says what worked there, simulation says whether it works here. The machine did not invent the courtyard, and it never learnt why the courtyard works.
Why we hold it
A regional passive type is the only environmental evidence that comes with real, occupied buildings attached — a climate and a street address. Correa built open-to-sky space into housing at Belapur and into a Bombay tower, and wrote down why. Baker's rat-trap bond and filler slab are in many Kerala houses that can be visited. An image model's "eco" is a pattern from photographs. A passive type is a precedent. That is the pattern-is-not-precedent distinction, applied to heat, and it is why the type comes first and the simulation comes second — as the test of the type on this site, not as its rival. The Lab's own notes on Baker and on critical regionalism sit behind this card.
The strongest objection
Precedent is evidence, not proof. "The vernacular did it" cannot establish that a courtyard will perform on this plot, in this geometry, at this density, for this household, in a city hotter than the village the type was built for. A jaali, a thick wall or a court is not passive-good in itself; its performance depends on orientation, proportion, occupancy and the hour, and a court on a six-metre plot can be an oven. Starting from the type can become starting from nostalgia, with "passive first" an excuse to skip the active systems a modern building needs and the measurement that would have shown the type failing.
What would make us revise it
We accept the objection's premise, which is why the position puts the simulation directly behind the type and not behind the generated scheme. We would revise the order if measured evidence, on Indian sites, showed schemes reached through generated imagery and simulation beating the regional type on comfort and energy without starting from it — the machine finding a move the repertoire did not have. We would also revise if the tools began to carry passive logic themselves: a generator that places a court because the weather file called for one. We look for both every edition.
Try it
Draw the plan of one old house of your own region's type from a visit — the court or the inner court, the verandah depth, the wall thickness, the roof. Then brief an image tool with those three facts and "no green wall, no solar panels." Compare what it gives you with what it gave you before. Write down what the brief had to contain before the picture started to look like a building.
Take it to crit
Ask the student which environmental moves in the scheme are passive — orientation, shade, mass, ventilation, the court — and which were supplied by the image tool's idea of "eco." Then ask them to strike the generated features on paper and say whether the house still works. If it does not, the scheme had no environmental design in it, only decoration.
How it works
A passive building manages heat without fuel. It keeps the sun off the skin (orientation, shade), slows heat through the wall (mass, a cavity), lets heat out at night (the court open to the sky, cross-ventilation) and filters light and air (the jaali). These are physics, and a simulation can show them working — or not working, on this site, which is the point of running one. An image model has none of this. It has photographs, and in photographs "eco" is a look. So the order of work matters: the regional passive moves first, because they cost nothing to run; then the simulation that tests them here; then, if you want one, a picture. That order is why the surrogate and simulate-then-trust cards come after this one.
What this idea builds on
What this idea opens up
Sources
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.